n8n vs Make vs Zapier – which fits Swiss SMEs?

What this is about

n8n, Make (formerly Integromat), and Zapier are the three most-used workflow-automation platforms in the DACH region. All three pursue the same goal: orchestrate business processes without coding between SaaS tools, databases, APIs, and increasingly LLM endpoints. Despite the same promise they differ fundamentally in licence model, hosting options, billing unit, and fit for regulated Swiss work.

n8n is a Berlin-built open-source tool (Fair-Code licence). Self-hosting is officially supported, the cloud variant optional. Make is a Czech SaaS product, hosted only, with a visually appealing scenario UI and operations-based pricing. Zapier is the US market leader since 2011, cloud-only, with over 7000 app integrations and task-based billing. As of May 2026 the picture is clear: n8n has AI nodes integrated natively, Make has refined its operations billing further, Zapier remains the most expensive but most compatible vendor.

Why this comparison?

A Swiss fiduciary office, a law firm, or an insurance agency does not ask "whether" to automate but "which platform". The choice has three hard consequences.

First data protection. n8n self-hosted can run on Hetzner Falkenstein or Exoscale Zurich – client data never leaves EU/CH. Make hosts in Prague (EU) but every data flow crosses Make servers. Zapier hosts in the US under standard contractual clauses; for data under professional secrecy (Art. 321 SCC) this is a no-go for many firms.

Second price structure. With 50 clients and 4 workflow runs per client per month (200 runs) a Swiss fiduciary pays roughly: n8n self-hosted CHF 0 (server only), n8n Cloud Starter EUR 24, Make Core from USD 10.59 for 10000 operations, Zapier Professional USD 73.50 for 2000 tasks. The gap multiplies with growth.

Third lock-in. Workflows from Make or Zapier do not export 1:1. Switching to n8n means re-implementation. Anyone starting with n8n can move between cloud and self-host at any time – same JSON workflow definition.

This comparison gives Swiss SMEs a clear recommendation per business type instead of an "it depends" answer.

Head-to-head on 6 axes

Data protection / Swiss DPA fit. n8n self-hosted clearly wins: no third-country transfer, own server, own logging. n8n Cloud hosts in Frankfurt (EU). Make hosts in Prague (EU), offers a GDPR DPA. Zapier hosts primarily in the US, DPA with standard contractual clauses and transfer impact assessment needed.

Price per 1000 runs (May 2026). n8n Self-host: about CHF 0.01 (electricity + server share only). n8n Cloud Starter EUR 24/month for 2500 executions = EUR 9.60 / 1000 runs. Make Core USD 10.59/month for 10000 operations – but caution, an "operation" is a single step, not a workflow run; real cost about USD 1.06 per 1000 runs at 10-step scenarios. Zapier Professional USD 73.50/month for 2000 tasks = USD 36.75 / 1000 tasks (task = one step). On a 10-step workflow Zapier is 30 times more expensive than Make.

Maturity / app integrations. Zapier leads with 7000+ apps, covers every niche tool. Make has about 1500+ apps, very good for the marketing stack. n8n has 600+ official nodes plus an HTTP node + code node for everything else; the AI nodes (OpenAI, Anthropic, Mistral, Ollama, LangChain) are richer than Make or Zapier.

Lock-in. n8n: workflows exportable as JSON, can be fully self-hosted, source open (Fair-Code licence). Make: no export outside the platform; replication on another tool = rebuild. Zapier: same, plus US vendor risk under FISA 702.

Self-host capability. n8n: yes, officially, Docker Compose or Kubernetes, one day of work. Make: no. Zapier: no. Anyone needing self-host for FINMA-related or legal workflows has only n8n as an option.

DACH fit. n8n: German-speaking forum, Berlin company, GDPR by design. Make: Czech/English, good DACH adoption in marketing. Zapier: pure English, US mindset; German support only on premium tiers.

Decision path in 6 steps

1. 01Data classification. Are client, patient, or HR data involved? If yes, mandatory candidate n8n self-hosted. If no, continue.
2. 02Estimate volume. Count workflow runs per month, multiplied by steps per run. Under 100 runs/month: Zapier or Make. Over 1000 runs: n8n.
3. 03Check hosting ability. Is a sysadmin or provider available for Docker/Linux? If no, n8n Cloud (EU) or Make.
4. 04Match app library. List required integrations and check against n8n nodes, Make apps, Zapier zaps. If a standard app is missing: check if HTTP node + API call is enough.
5. 05Project costs over 12 months. n8n self-host: only server cost; cloud vendors: pick the tier with a 20 percent buffer.
6. 06Build a pilot. Implement one productive workflow on the favoured tool, measure 2-4 weeks, then scale.

When which platform

When n8n. You are fiduciary, lawyer, doctor, or insurer. You handle client or patient data. You want self-host on Hetzner or Exoscale. You build LLM-backed workflows (document recognition, email triage, client onboarding) and want no lock-in. You have a sysadmin or provider who can read Docker Compose. n8n self-hosted is the default for regulated Swiss SMEs.

When Make. You are a marketing agency, e-commerce shop, or operations team with main toolchain in HubSpot, Klaviyo, Shopify, or Notion. You want fast start, no server. You have no data under professional secrecy. Your workflows have many steps but volume is mid-range (10k to 100k operations/month). Make has the best price per step in this volume class and the nicest UI.

When Zapier. You need exactly one very niche integration that only Zapier covers (example: QuickBooks Desktop, old Salesforce custom objects, niche CRM). Your volume is under 100 tasks/month. You have no hosting know-how. The user experience must be one-click-done and price is secondary. For everything else, Zapier in 2026 is too expensive.

When NONE. If your workflow needs thousands of steps per run or sub-second latency, none of the three is ideal – then build a Python service on your own infrastructure. If your client data is under FINMA supervision and absolute tier-A sovereignty is required, even n8n self-hosted is only one building block – compliance is in the setup, not the tool.

When NONE of the three

If you only need one or two workflows with two steps each ("email in, log to sheet"), a 5-line Python script is cheaper and more stable. If you build multi-step LLM agents with memory and tool use, LangChain or LlamaIndex is the better choice – the three workflow tools have AI nodes but no real agent framework. If you need real-time latency under 200 ms (voice agents for example), all three are too slow – go direct-server with webhooks.

Additionally: if you already have an established ESB (MuleSoft, IBM ACE), a workflow tool only adds complexity. If you only sync data between two Postgres databases, a pure ETL solution (Airbyte, dbt) fits better. If your compliance brief requires absolute data sovereignty including the logging pipeline (security-cleared industries), you need a solution whose full audit trail you control – even n8n self-hosted alone is not enough without additional Loki/Grafana configuration.

Trade-offs

FAQ

Related topics

Sources

1. n8n.io – Cloud and self-host pricing · 2026-05
2. Make.com – Pricing plans · 2026-05
3. Zapier – Plans and pricing · 2026-05
4. Toolradar – Zapier vs Make vs n8n cost analysis 2026 · 2026-04
5. n8n docs – Self-hosting and Docker Compose · 2026-05